OpenStack Infrastructure as Code

Terraform configurations and automation for deploying and managing OpenStack cloud infrastructure.

Overview

Infrastructure-as-Code modules for OpenStack environments, developed from real-world experience managing OpenStack across multiple datacenters.

Experience Context

• Deployed OpenStack across 7 datacenters using Kolla-Ansible
• Managed CEPH distributed storage at petabyte scale (90+ nodes, 6 DCs)
• Achieved 35% RAM reduction on CEPH clusters, saving $150K+ in hardware costs
• Reduced datacenter bandwidth costs by 65% through traffic optimization

Tech Stack

• Terraform (HCL)
• OpenStack APIs
• Kolla-Ansible
• CEPH Storage

Structure

Infrastructure modules and configurations for OpenStack resource provisioning and management.

Deployments

• cloudinative-openstack — Kolla-Ansible 2026.1 (Galaxy) all-in-one on a nested-KVM VM (24 vCPU / 96 GB / 1 TB), full service set incl. Octavia, Designate, Magnum, Skyline. Patches + Dockerfile overlays + systemd units + post-deploy bootstrap script — reproducible from a fresh KVM host. Also documents the planned migration off the dead k8s_fedora_coreos_v1 Magnum driver to magnum-capi-helm.

Guides

• Recovering a Kolla-Ansible Galera split-brain — and the /etc/hosts gotcha that caused it — full walkthrough of diagnosing and recovering a 3-controller Kolla 2025.1 cluster after Keystone went 503, including the cloud-init manage_etc_hosts failure mode that re-arms the same outage on every reboot.
• Locked out of an instance: fixing SSH by editing its root volume offline — proving whether a refused port is the client, the security group or the guest; mapping a Ceph RBD root volume to get a shell on a VM you cannot boot into; and the two root causes it surfaced — a ssh.socket drop-in that listened on IPv6 only, and TLS material bind-mounted from /tmp.